Best practices for handling replication transition

Posted by Jens Dahlin on 29-Nov-2017 04:45

Are there any best practices for taking care of clients in a mixed (webspeed, appserver, client networking, batches) environment when doing a transition from one replicated server to another?

Nameservers can be used to tell a client what broker to connect to but what are the options for how to decide what database host to connect to? A local DNS? Other options? 

Also brokers needs to be at least refreshed, possibly restarted for changes to take effect?

All Replies

Posted by lbryan on 07-Dec-2017 09:12

When I have set up OE replication in the past I have used a DNS entry for the database server, therefore if the source is failed over to the target, this DNS entry can be re-assigned to the server that has been promoted. This makes it much easier for any configuration that you have the references the -H parameter. None of these have to be updated as they are still looking for the same name.

You would still need to reconnect any connections to the database as during failover the source database shuts down(whether planned or not). Not sure how it works with webspeed but for appserver, client networking and batches the connections would need to be remade. Usually for app servers this would entail a restart, I did come across some additional code that looks as though it makes an automatic app server reconnect in but I have not implemented this - if you do I'd be interested to hear how you got on.

Client networking connections would need to be remade, batches that run on a schedule should be ok provided that they too connect using the DNS name.

Anywhere that you had the I.P address or the machine name imbedded would fail to connect after failover.

Hope this helps.

This thread is closed